Convert GB to kB

A gigabyte is a billion bytes in the decimal sense used by drive manufacturers, phone plans and video sizes.

A kilobyte is 1,000 bytes when a standards body defines it and 1,024 when an operating system does. This page uses 1,000; the kibibyte is the unit that means 1,024.

Going from gigabytes to kilobytes moves the decimal point 6 places to the right and changes nothing else. There is no factor to remember and no rounding to decide: the digits stay in the same order and only their position changes.

1,234 GB is 1234000000 kB — same digits, moved along. That is worth knowing because it is the one kind of conversion you can check at a glance: if the digits of the answer are not the digits you started with, something other than the conversion has happened to them.

Six places of decimal point separate the two units, and nothing anybody would call a file sits at both ends of that gap. Kilobytes are a unit price — the cost of one request, one page view, one log line, one telemetry sample. Gigabytes are a budget. Converting downwards turns the budget into the currency it will actually be spent in, which is why the answer is nearly always a count rather than a size.

That count is the form worth writing down. A gigabyte is a million one-kilobyte requests, twenty thousand fifty-kilobyte pages, two thousand five-hundred-kilobyte images. Each of those is a sentence somebody can check against what they expect to happen in a month, and none of them requires holding a seven-digit number in the head.

Divide the million by the per-item size and the answer arrives directly: at 40 kB a page view a gigabyte buys 25,000, at 400 kB it buys 2,500, at 4 MB it buys 250. The arithmetic is trivial and the spread is the point — a factor of a hundred between the lightest and heaviest realistic per-item cost means the same allowance can be enormous or negligible depending on one number nobody measured.

The per-item figure is therefore the thing to get right, and it should be measured rather than assumed. A page-weight tool reports what a page transfers on a cold load; a repeat visit costs far less because most of it is cached; a logged-in view may cost far more. Where a quota matters, the useful measurement is a real session over a real connection rather than a single number taken from a build report.

A plan sold as a number of gigabytes is spent in kilobyte-sized pieces, and the pieces are mostly not the ones a user chose. A single article page can carry fonts, several tracking scripts, an advertising stack and autoplaying video, and the share of the transfer that is the article itself is often under a tenth. Converting the allowance to kilobytes and dividing by a realistic page cost gives a count that explains why a month runs out in a fortnight.

The largest single line item is usually video, by an order of magnitude that no amount of page tuning approaches. Standard-definition streaming spends roughly a gigabyte an hour and high definition several times that, so an hour of video is worth thousands of text pages. When an allowance disappears without explanation, the arithmetic almost always points at a video that played without being asked to.

A structured log line with a timestamp, a level, a message and a handful of fields is commonly between 200 bytes and 2 kB, and the whole difficulty of log budgeting is that the count is the variable, not the size. A thousand lines a second at 500 bytes each is 500 kB a second, which is 43 GB a day — a figure that arrives without any single decision having been taken to spend it.

Converting the budget downwards makes the constraint legible: a 100 GB monthly allowance is 100 million kilobytes, which at 500 bytes a line is 200 million lines, which is about 77 lines a second sustained. That last number is one an engineer can compare against a log level, and it usually shows that the choice between INFO and DEBUG is the whole budget rather than a preference.

A budget over a period is a rate, and rates catch things that totals hide. A gigabyte a month averages about 386 bytes a second; a gigabyte a day is roughly 11.6 kB/s. Anything continuous — a heartbeat, a metrics agent, a polling loop — should be checked against those figures, because a process sending 2 kB every ten seconds is spending half a gigabyte a month and looks entirely innocuous in isolation.

The reverse conversion is equally useful when reading a specification. A device advertised as using "a few kilobytes an hour" is under a hundred megabytes a year and can sit on the smallest plan available; one using a few kilobytes a second is tens of gigabytes a year and needs a real allowance. The unit of the numerator and the unit of the denominator both have to be read, and it is the denominator that people skip.

Moving a decimal point six places is where an order of magnitude quietly goes missing, because 1,000,000 and 10,000,000 look alike at a glance and both are plausible answers. Counting in named steps rather than in zeros is more reliable: a gigabyte is a thousand megabytes, a megabyte is a thousand kilobytes, so a gigabyte is a thousand thousand kilobytes. Two steps, each of which is easy to check.

The mistake survives longer than it should because the wrong answer is still a reasonable-looking number. A quota that is out by a factor of ten does not fail immediately; it fails at the end of the month. Running the arithmetic back — multiply the count of items by the per-item size and see whether it returns the original gigabyte figure — takes a few seconds and catches exactly this class of error.

Networks, providers and bills are decimal: a gigabyte of transfer is a thousand million bytes, and the meter on the far side counts bytes rather than anything friendlier. That is convenient, because it means the conversion to kilobytes on this page is exactly the arithmetic the provider is doing, and no adjustment is needed between the allowance and the count.

The discrepancy appears on the device instead. An operating system reporting per-application data use may divide by 1,024 three times, so its gigabyte is 7.4% larger than the network’s and its tally of the same month comes out lower. A user comparing the two concludes that somebody is overcharging. Where the difference matters, both sides can be reduced to bytes, and the byte counts will agree.